{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.09450", "kind": "paper", "url": "https://arxiv.org/abs/2606.09450"}], "authors": ["QuocViet Pham", "Elvir Karimov", "Andrey Galichin", "Ivan Oseledets"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.09450", "source_url": "https://arxiv.org/abs/2606.09450"}, "categories": ["Language & Knowledge", "Mathematics & Formal Science", "cs.AI"], "description": {"en": "Evaluates LLMs on theorem proving in Lean4 using classical theorems, with two versions: main and premised. Includes metrics for theorem-level coverage and token efficiency to assess partial progress and proof structure."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.09450", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.09450", "title": "TheoremBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.09450", "modality": null, "name": "TheoremBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_theorembench_8bd203fd", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "2d7925ed1caae0529d80c15ac83280f396d6682c80eedb9b24d231124c0494ce", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.09450"}, "publisher": null, "released": "2026-06-08", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.09450", "source_url": "https://arxiv.org/abs/2606.09450"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-09450", "source": "claire_radar", "source_benchmark_id": "2606.09450", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Mathematics & Formal Science"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.09450", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "unavailable"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-21T04:30:40.934319Z", "inputHash": "d903f06021d84a12e5ccc9bb1a83178e23d4f6730d1725d9b4118f5c72bd16ad", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates LLMs on theorem proving in Lean4 using classical theorems, with two versions: main and premised. Includes metrics for theorem-level coverage and token efficiency to assess partial progress and proof structure.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce TheoremBench, a Lean4 benchmark designed to evaluate theorem provers beyond contest settings."}, "familyId": "bmf_e8410a03abbd", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_theorembench_8bd203fd", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.09450", "paper": "https://arxiv.org/abs/2606.09450", "pdf": "https://arxiv.org/pdf/2606.09450", "project": null, "report": "https://arxiv.org/abs/2606.09450"}, "metrics": [], "motivation": "LLMs have recently achieved strong results on formal proving benchmarks.", "name": "TheoremBench", "oneLine": "Evaluates LLMs on theorem proving in Lean4 using classical theorems, with two versions: main and premised. Includes metrics for theorem-level coverage and token efficiency to assess partial progress and proof structure.", "paperTitle": "TheoremBench: Evaluating LLMs on Theorem Proving in Formal Mathematics", "primaryDomain": "Mathematics & Formal Science", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.09450", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "TheoremBench Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://arxiv.org/abs/2606.09450"}], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-08", "source": {"authors": ["QuocViet Pham", "Elvir Karimov", "Andrey Galichin", "and Ivan Oseledets"], "categories": ["cs.AI"], "id": "2606.09450", "title": "TheoremBench: Evaluating LLMs on Theorem Proving in Formal Mathematics", "type": "arxiv", "url": "https://arxiv.org/abs/2606.09450"}, "sourceUpdatedAt": "2026-06-08T12:57:18Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Provides a more realistic evaluation of provers beyond contest problems, revealing biases toward easy subtheorems and inefficient proof strategies. Supports finer-grained analysis of formal reasoning capabilities."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
